In response to the much touted prospective gains in pharmacology that AI would bring (see DeepMind's Alphafold), this writer argues that, to the contrary, AI's achievements will be limited by the number of remaining drug targets.
Drug targets are "the large molecules that drugs bind to. Nearly all of them are proteins, which are of course encoded by genes. There aren’t that many human drug targets (bacterial and viral targets are a different story). The human genome encodes only 19,000 proteins. About a tenth of these are implicated in disease and are plausible drug targets. About half have already been “drugged”." Basically, the argument goes that AI will be unable to find what isn't there.
On the plus side, according to this theory, AI will mostly help in discovering drugs for niche diseases or perhaps solve "some longstanding conundrums in drug development. A non-toxic gain-of-function ligand for the tumor suppressor protein p53, often described as “undruggable”". Not enough to save the loss-making pharma industry, but certainly of value to those suffering from these diseases.
Read here.
